If you spend a lot of time in the shell and are used to manipulating text via chains of piped Unix utilities, then MPD controlled by mpc lets you put music on without leaving that environment that is comfortable and efficient. For example, play album X but not tracks Y or Z from it that I think suck. Or play such-and-such a piece of classical music but in the recording by that particular conductor and ensemble, etc.

EDIT: The below is completely wrong, ignore.

<del>The server-client model means you can run the server (MPD, Rockbox Zig, whatever) on your homelab, and stream music to multiple clients. So you don't have to copy all your music onto your phone, you can just run an MPD client like M.A.L.P. and it will stream from the files on your server. Then there's just one copy of your music collection, tags, etc.: no more keeping tags up-to-date between the copies on your home server, your laptop, your phone... (Do make sure your files are backed up, of course).<del>

Our IT department has a machine called jukebox hooked up to a sound system. Since it runs a music player daemon everybody in the IT department can control the jukebox from their own frontend, phone or whatever.

Technically you could achieve the same with a web frontend, but then you're limited to that UI. A music player daemon is agnostic to the UI, meaning there can be different UIs or you can add music via command line, text interfaces etc.

> Rockbox Zig exposes the full Rockbox audio engine — gapless playback, DSP, 20+ codecs, tag database — through gRPC, GraphQL, HTTP, and MPD APIs

I like having the music player be separate from the client that controls it. I'm currently listening to music on my home desktop PC while using my work laptop, and I can control the music playback from the work laptop by just connecting to the daemon.

You can also use multiple clients if you want. Some TUI, some graphical, some utility like mpdscrobble (that just watches what you listen to and scrobbles it to Last.fm).